Roche: Boosting Case Law Analysis With Generative AI

European patents in the pharmaceutical sector often lead to oppositions and appeals and require extensive legal analysis and understanding of the European Patent Office (EPO) Board of Appeals case law, which encompasses approximately 40,000 cases. The process currently demands a massive investment of patent attorney work hours, involving time-consuming and costly searches through reference books and keyword-based research.

Process Hurdles: Inefficiencies in Case Law Analysis

The existing approach to case law analysis is inefficient and incomplete. Patent attorneys invest considerable time conducting manual searches and analyzing complex legal documents to prepare for oppositions and appeals. This labor-intensive process not only consumes valuable resources but also limits the ability to generate insights quickly.

The lack of real-time data processing and the reliance on manual search significantly hampers the team’s effectiveness and creates a need for a more streamlined solution.

From Manual to Automated: Generative AI-Driven Case Law Analysis

Recognizing the potential of Generative AI and the future of advanced analytics, Roche patent attorneys turned to the power of Dataiku. A pilot project, dubbed “chatWhitebook,” was established to facilitate semantic search and full-text analysis of relevant appeal cases. Utilizing citizen development and GPT-4, the team of patent attorneys developed a tool that provided essential insights directly to patent attorneys, significantly improving the efficiency of case law analysis.

The implementation involved several key features of Dataiku, including the LLM Mesh, Knowledge Bank, and Prompt Studios, alongside visual recipes for data processing. Together, these were used to create a Dataiku application to execute a complex search and retrieval flow for comprehensive case law analysis. Extensive testing was conducted with experienced patent attorneys, ensuring the system delivered accurate and actionable insights tailored to their queries.

Evaluating different prompts and estimating token costs using/comparing different LLMs was easy on the Dataiku platform. Andres Buser Chapter Lead New Modality IP, Roche

After the chatWhitebook flow was fully operational on the Dataiku platform, plans sprouted for a broader rollout across Roche’s patents department.

Catalyzing Efficiency: The Long-Term Benefits of AI Integration at Roche

With Dataiku, we've demonstrated that citizen development can rapidly build complex AI solutions. This approach has allowed us to initiate projects that might have otherwise been delayed by traditional processes, ultimately enhancing our capabilities in case law analysis and beyond. Andres Buser Chapter Lead New Modality IP, Roche

The integration of Dataiku into processes of Roche Group Patents marks a significant shift in operational dynamics. With citizen development, Roche patent attorneys are enabled to quickly build experimental use cases without the need for extensive IT support. This newfound agility fostered an environment where legal experts could focus on high-value tasks rather than tedious data processing.

The entry barrier to starting with AI and data analytics tools is remarkably low when using the Dataiku platform, making it accessible even for beginners/non-coders. Using the Dataiku platform and leveraging self-service or citizen development, one can achieve rapid results without the need for additional IT experts. Andres Buser Chapter Lead New Modality IP, Roche

Roche has confidently positioned itself to harness deeper insights from court cases, potentially increasing its success rate in opposition and appeal proceedings. By streamlining case law analysis, the company has estimated annual savings between $100,000 and $250,000, primarily through reduced attorney hours per case.